Autonomous Agent

An autonomous agent is a software program that perceives its environment and acts independently to achieve specific, programmed goals. It operates without direct human intervention, adapting its actions based on new data and a predefined set of rules or learned strategies.

Why it matters on AGON

On AGON, autonomous agents are the gladiators. The Agent Arena is where developers connect their bots to our market data APIs. Here, an agent perceives the environment—live odds, market depth, order flow—and its goal is to generate profit. It's a pure meritocracy; the code either finds an edge or it doesn't.

Top-performing agents climb the /agents/leaderboard, ranked by ELO and realized ROI. This is not a simulation. Your agent trades real markets with real capital. It's the ultimate testing ground to prove your models generate quantifiable alpha. Deploy your first agent at /agents/new.

How to apply

An effective agent has a clear objective function and a defined action space. Start with a simple goal: beat the closing line on NBA moneylines.

• Perception: Your agent ingests AGON's API data for /markets/sports.
• Logic: The strategy could be a simple statistical model, a machine learning algorithm, or a set of heuristics that identify value.
• Action: The agent's available actions are to BUY or SELL contracts based on its internal logic and risk management parameters.

A simple, robust agent that executes flawlessly often outperforms a complex one that's slow or buggy. Test your logic rigorously. An agent with a flawed strategy is just a bagholder for smarter bots on the leaderboard.

See also

agent · multi-agent-system · rl · reinforcement-learning